ニューラルネットを用いた顔画像識別の一検討 : 視聴覚技術

抄録

The back propagation network,(BPN) is applied to human-face identification. A mosaic pattern with 12 X 12 blocks, transformed from central part of human-face image, is put into BPN. This combination yields hundreds of person's identification with robustness to deformation such as deforcused image or different face expression. Hidden units of the BPN extract peculiar and delicate features of human-face, which can not be obtained from existing statistical methods. A few of hidden units can especially select only men or women.